Feeding SMPS and other Electronics loads from Inverters 2

[Grollor33]

Hey all,

I understand all designs are a little different...I'm just looking for generic guidance....thanks a lot!!

I have a situation where I need to feed many (what of I'm guessing are)switch mode power supplies along with HMI displays, video cameras, various communication devices, etc. from power inverters/UPSs.

Power ranges from 5kVA to 250 kVA inverters...each with a mix of the "electronics" loads.

One of the main issues that has been raised is the startup of the system. My questions are for folks familiar with both (either) sides of the power system: Power Inverters & SMPSs

SMPS Question: What do these devices actually look like when they start up? Will they pull a slug of inrush current in order to charge up their input caps? If so, is it purely due to dV/dt on the cap? Is there a minimum voltage where the unit won't pull any current (during startup) and then "instantaneously" pull the starting current? Will the inrush be over within a cycle or two?

Inverter Question: We've spec'd the inverter to current limit to 2x its rating. Does anyone have experience in feeding these electronics' inrush currents? I'd like to think that I'm worrying about nothing and the regulation of the inverter will allow a cycle or two of high current to pass thus supplying the charging currents without placing the unit into a current limited state. Is this anywhere near the reality of what will happen?

There's also been worries of breaker coordination with a current limited source. Anyone have experience with feeding distribution circuits from power electronics components? Did you have the same nightmares that I'm having?

Thanks a lot.

 

[Skogsgurra]

There are a few standard breeds of switch mode power supplies. Som have a diode bridge and a capacitor - not much more. Other have OFC input circuits and some have brown-out protection.

The first type will definitely take a large inrush current. Ther's not much to limit the current.

The second type draws a current wave-shape that is a copy of the voltage waveform. The thinking is that if the voltage waveform is a sine, then the current will also be a sine and the power factor will be very close to unity.

The brown-out protection typically switches a SMPS off when voltage gets too low. And it (probably) doesn't activate the supply if voltage hasn't reached a certain, minimum value.

If there are frequency inverter loads, they usually, but not always, have some means of limiting inrush current.

The UPS can also have its problems with these loads. The term "Electronic Instability" has been coined to describe different problems that can occur when combining different loads and UPSes.

I think that these and other problems like protection coordination etc you will encounter makes hiring of a specialist in this field a good idea.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[IRstuff]

As a general rule, you don't try to bring up EVERYTHING all at once. Even if you don't pop a breaker somewhere, you'll likely overdraw the source.

Typical military systems power up only critical systems initially, then bring up secondary and tertiary systems.

TTFN

Eng-Tips Policies FAQ731-376

 

[Skogsgurra]

Hmm... OFC shall, of course, read PFC. Sorry for any confusion.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Grollor33]

Thanks Skogsgurra for the insight into the SMPSs. It seems that I need to do some more digging concerning exactly what type of SMPSs that I'll be feeding.

IRStuff, this is very similar to the military application you describe. Are you aware of any mil-standards or legacy designs which would have this type (or similar type) of power system that I could use for guidance?

 

[jimkirk]

You might consider specifying the inverter as something like
x1 % of rating for y1 seconds (or milliseconds)
x2 % of rating for y2 seconds
x3 % of rating for y3 seconds
to allow very high inrush currents for short times while not overspecifying the whole thing. Inrush can easily be 10 times normal loads.

As has been said, different switch mode supplies have different characteristics, and you might want to sequence turning things off.

I'd recommend taking a similar look at how you turn things off. Turning off lots of loads all at once may cause transients that could damage the inverter or loads that are still connected.

 

[IRstuff]

I don't think there are overt requirements. However, reading of MIL_HDBK-454, MIL-STD-1275B, and MIL-STD-704F would give you a flavor that there is an expectation of 10x surge current vs. steady state, but that the surge currents are not allowed to cause the power source to violate 704 or 1275, whichever is applicable.



TTFN

Eng-Tips Policies FAQ731-376

 

[jraef]

Hmm... OFC shall, of course, read PFC. Sorry for any confusion.

Glad you cleared that up. I thought maybe it meant Old Fart's Control, something I thought you'd be familiar with Gunnar!

 

[BobbyNewmark]

I wouldn't rely on what kind of SMPSes are used. Few vendors will sign up to guarantee what they use in their equipment[—]they might not even know themselves. It is also likely to change without warning, as products are redesigned to reduce cost. Assume the worst and build for it.

For staged power up, consider using electronic switches that close during zero crossings (when the AC line voltage is near zero). That dramatically reduces the surge current needed to charge up capacitors.

 

[bogeyman]

Beware of some auto voltage selecting SMPs that will detect a 100v supply while the invertor gets going and then carries on going until it gets up to 240v. We lost 6 units in PCs before some one at the invertor makers twigged what was happening.

 

[Grollor33]

Thanks for all the insight thus far...

So from what I've read so far, the safest approach is to assume worst case SMPS design...which, as far as inrush is concerned, is diode bridge-cap combo without out any type of limiting inductive reactance.

My assumption is that the inrush event is over within a few electrical cycles (1-2 at 60 Hz). Good assumption?

I have nearly 1500 of these loads being sourced from less than 100 inverters...with the electronic switches, it seems like the cost could easily get out of control - even though the sophistication and granularity of control of it is desirable. There's little doubt that I'll have to sequence blocks of loads. Unfortunately we already have a design which has the 2x current limit and its significant cost to change it.

Bogeyman, could you expand on the problem you encountered?

 

[BobbyNewmark]

100 inverters?! Is there any possibility of simplifying it into a few big inverters?

Do you have any loads that draw lots of start-up current? Many CRT-type computer displays have a degaussing coil that activates when the power turns on, drawing a moderate current for several seconds. Then there are laser printers and photocopiers, which draw high current for tens of seconds. Big hard drive arrays can also take a lot of power to spin up.

 

[Grollor33]

We're stuck with 100...

Hard drive arrays are what a lot of the load will be...monitors and a few printers/etc...

It scare's me a bit when you say "a lot"! Breaker coordination is also ugly with a limited source that supplies loads that want "a lot" of power during startup but not that much during SS operation...

I guess the issue is not an imagined one (as I was hoping)...Thanks for the reply!

 

[BobbyNewmark]

It's definitely a real consideration.

I think this is a solved problem for hard drive arrays. The hard drives can be indivdually spun up and down under software control, and I've heard that many array controllers can be told to stagger drive start up. (In fact, el cheapo arrays HAVE to, otherwise they overload their internal power supplies.) With luck, this part will practically solve itself.

CRT degaussing is a moderate problem. The surges are generally just a few amps, and if you get painted into a corner you can throw LCDs at the problem areas. Beware of a bank of monitors plugged into one power strip: somebody will eventually cycle the strip's power button.

Copiers, laser printers, and LED printers are the real bear, well known for overloading inverters and UPSes. They use a massive heater to melt the toner to the paper. Some models draw over 10 amps when the heater is warming up, and the current is rather unpredictable, depending on workload, ambient temperature, and power saving modes. If somebody split a big print job across ten printers for speed, you might well see the load instantly increase by 20 kilowatts, and keep going for tens of seconds, as all the printers come out of power-save mode. The usual solution is to limit the number of printers/copiers that must run on inverter power, and get stout inverters for the ones you can't live without. Read inverter and printer specs carefully, and be sure to run your qualification tests with cold printers to get true worst-case results.

 

[ScottyUK]

What is this inverter powered from? Any chance of using a COTS UPS module instead? All the debugging will have been taken care of and these are designed to withstand surges etc caused by computer-type loads. As far as the inverter is concerned, a laser printer or similar should be considered as a continous load.


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[Grollor33]

COTS units are typically designed to withstand these surges??

I've been surfing around trying to find COTS unit's specs that speak to anything regarding the startup transient. I haven't found anything really specifed...at least not where I've been looking...It seems like this is just one of those issues that everyone knows about but isn't really spoken about. Any companies out there that you know of that could provide such a unit?

 

[ScottyUK]

Looking at a fairly large GE Signature UPS just because the manual happens to be on my desk, and picking out some of the salient points:

The inverter stage short circuit current is limited to 2.2x nominal rating for 200ms. It will handle an overload of 150% for 1 minute and can handle a crest factor of greater than 3:1. Nothing in there is particularly unusual - if I'd picked a module from any of the other major manufacturers it would be very similar.

You are right to be concerned about breaker coordination. The best way of dealing with this is to use many small individual circuits. To give you some idea, an 80kVA module won't reliably clear anything much bigger than a 16A or 20A Type 'B' MCB.

What's your DC supply - a battery? What voltage?


----------------------------------

Sometimes I only open my mouth to swap feet...

 

[Skogsgurra]

A solution to this kind of problem might be a very neat - and very old - technology called DC.

There are a few test installations running here where the whole office is fed from a 350 V DC system.

Benefits:
* One large "UPS" (a simple battery, actually).
* Some installations may even be able to use super capacitors instead of batteries.
* More power per sqmm Cu (higher voltage and peak voltage = RMS voltage), which translates into less losses at a given conductor size.
* Very simple equipment. A rectifier with its control circuitry.
* No harmonics. (DC has no harmonics).
* No HF pollution from-less-than-well designed UPSes. (Yes, some of them are major sources of EMI).
* Reduced loss computer's, drive's, printer's etcetera power supplies. Intel has an interesting paper on that.
* Reduced loss means less waste heat removal - a strong CO2 statement.
* Reduced maintenance. One common battery to look after instead of hundreds of small ones. How many UPSes do work when needed? Honest?

France Telecom has recently received a test system for their new HiTec building in France. Nippon Telecom and NEDO looks at this technology. And probably a lot more people that I haven't met or heard about.

Google "DC distribution", UPN, NetPower etc for more ideas.


Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Grollor33]

I wish I could have my users convert to all DC...especially since that what they usually end up doing with the AC that I provide them anyway!


With the exception of the 2.2 pu current limit, thats basically what I've seen out there. So would you say that the 150% overload is where the starting transients are supplied from?

 

[ScottyUK]

Anything which lasts a minute isn't a transient - that 150% is the short term overload capability. Are you worried about the inrush which lasts maybe a couple of cycles when the SMPS input stage charges up or longer duration heavy loads such a a laser printer spinning up?

Skogs mentioned the easiest way out for the first problem: use loads with a PFC active front end. No inrush or distorted waveforms.


----------------------------------

Sometimes I only open my mouth to swap feet...